https://github.com/scutbioinformatic/causalcall

https://github.com/scutbioinformatic/causalcall

Last synced: about 2 months ago
JSON representation

• Host: GitHub
• URL: https://github.com/scutbioinformatic/causalcall
• Owner: scutbioinformatic
• Created: 2019-11-27T14:53:33.000Z (over 4 years ago)
• Default Branch: master
• Last Pushed: 2019-11-28T07:40:11.000Z (over 4 years ago)
• Last Synced: 2023-10-20T01:52:53.494Z (9 months ago)
• Language: Python
• Size: 37.5 MB
• Stars: 9
• Watchers: 4
• Forks: 3
• Open Issues: 2
• Metadata Files:
  • Readme: README.md

Lists

• awesome-nanopore - Causalcall - [Python] - [Causalcall: Nanopore Basecalling Using a Temporal Convolutional Network](https://www.frontiersin.org/articles/10.3389/fgene.2019.01332/full) (Software packages / Basecalling)

README

        
# Causalcall

Code for the paper *Causalcall: nanopore basecalling using a temporal convolutional network*.

# Environment

- Ubuntu 14.04 

- python 3.6 

- tensorflow 1.8 

Dependencies: 

numpy, collections, threading, tempfile, h5py, statsmodels, difflib, argparse, tqdm, multiprocessing, psutil

# Usages

### Preparing the training data:

Extract training and validation data (.tfrecords) from resquiggled fast5 files:

```

python raw.py -i training/validation_fast5_files_folder -o tfrecords_files_folder -f train.tfrecords/validate.tfrecords

```

### Training:

Train the model with default parameters:

```

CUDA_VISIBLE_DEVICES=cuda_id python train.py -i tfrecords_files_folder -o model_folder -m model_name

```

Make sure that train.tfrecords and validate.tfrecords are both in the tfrecords_directory.

If you want to change parameters, use `train.py -h` for more details.

### Basecalling:

```

CUDA_VISIBLE_DEVICES=cuda_id python basecall.py -i fast5_files_folder -o results_folder -m path_to_model

```

Path to default model:   *./model/DNAmodel/* 

### Acknowledgement

We thank Chiron for providing the [source code](https://github.com/haotianteng/Chiron). Causalcall is developed on the basic framework of Chiron's code.

(The parts of preprocessing input data and converting outputs of the TCN-based model to base sequences are revised based on Chiron's code following MPL 2.0.)